La serie de la temperatura fue extraĆda de la base de datos del IDEAM, se tuvieron en cuenta los promedios diarios de las temperaturas en grados centigrados (Ā°C) registradas en BogotĆ” en las diferentes estaciones meteorolĆ³gicas que recolectaron informaciĆ³n de esos dĆas. La serie de tiempo cuenta con un total de 1826 registros, de los cuales 13 (0.7%) fueron imputados puesto que no se presentaba la informaciĆ³n necesaria. Esta imputaciĆ³n fue realizada a partir del mĆ©todo de vecinos mĆ”s cercanos (KNN), donde se tuvierĆ³n en cuenta 5 vecinos.
Para el modelar la serie de la temperatura se usara inicialmente un modelo SARIMA, seguido de redes recurrentes simples, LSTM y finalmente LSTM. Para seleccionar el mejor modelo se tomarƔ como criterio el error cuadrƔtico medio.
import numpy as np
import pandas as pd
import plotly.express as px
import matplotlib.pyplot as plt
import sklearn as sk
from sklearn import impute
from sklearn import preprocessing
import sklearn.externals
import joblib
from sklearn.model_selection import TimeSeriesSplit
from sklearn.impute import KNNImputer
import sklearn.preprocessing
from keras.models import Sequential
from tensorflow.keras.callbacks import EarlyStopping
import time
import sklearn.externals
import joblib
import plotly.graph_objects as go
from sklearn import metrics
import statsmodels.api as sm
import statsmodels.tsa.stattools as ts
from tensorflow.keras.models import Model
from statsmodels.tsa.statespace.sarimax import SARIMAX
import tensorflow as tf
import tensorflow.keras.layers as L
import tensorflow.keras.models as M
import tensorflow.keras.backend as K
from tensorflow.keras.optimizers import Adam
%matplotlib inline
import plotly.io as pio
pio.renderers.default='notebook'
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
df = pd.read_csv("/content/drive/Shareddrives/Mineria /Temperatura1.csv", sep=';', header=0, decimal = ',')
Fecha = pd.date_range(start='2017-01-01', end='2021-12-31', freq='D')
df['Fecha'] = Fecha
df = df.set_index('Fecha')
print(df[pd.isnull(df.ValorObservado)])
print('En total hay' ,
str(df['ValorObservado'].isnull().sum()) ,
'valores sin informaciĆ³n')
print('Correspondientes al {:.3f}% del total'
.format(df['ValorObservado'].isnull().sum()*100/len(df)))
ValorObservado Fecha 2017-08-12 NaN 2017-12-24 NaN 2019-09-15 NaN 2019-09-16 NaN 2019-09-17 NaN 2020-11-12 NaN 2021-01-05 NaN 2021-01-06 NaN 2021-01-07 NaN 2021-01-08 NaN 2021-08-18 NaN 2021-08-20 NaN 2021-12-05 NaN En total hay 13 valores sin informaciĆ³n Correspondientes al 0.712% del total
La serie presenta valores faltantes, por lo tantol se imputaran usando el mĆ©todo de vecinos mĆ”s cercanos (KNN), como se muestra a continuaciĆ³n.
#ImputaciĆ³n de Valores usando el vecino mĆ”s cercano
imput = KNNImputer(n_neighbors=5, weights="uniform")
# Ajustamos el modelo e imputamos los missing values
imput.fit(df[['ValorObservado']])
df['ValorObservado'] = imput.transform(df[['ValorObservado']]).ravel()
print()
print("Valores pƩrdidos en ValorObservado: " ,
str(df['ValorObservado'].isnull().sum()))
Valores pƩrdidos en ValorObservado: 0
fig = px.line(df, x=df.index, y="ValorObservado")
fig.update_xaxes(title_text="Fecha")
fig.show()
Para el respectivo anƔlisis se tomarƔn el 80% de los datos para entrenamiento y validacion, el 20% restantes para prueba, dichos valores corresponden a 1460 y 365 respectivamente.
from sklearn.preprocessing import MinMaxScaler
# crea el objeto scaler y escala los datos
scaler = MinMaxScaler(feature_range=(0, 1))
scaled_data = scaler.fit_transform(df.values)
#
df_norm = pd.DataFrame(scaled_data,index=df.index, columns=['ValorObservadoNormalizado'])
train_size = int(len(df_norm) * 0.8)
test_size = len(df_norm) - train_size
train, test = df_norm.iloc[0:train_size], df_norm.iloc[train_size:len(df_norm)]
len_train = len(train)
len_test = len(test)
print(len_train, len_test)
1460 366
def create_dataset(X, y, time_steps=1):
# crea dos listas vacias para depositar los datos
Xs, ys = [], []
# el primer lote de datos empieza en la primera observaciĆ³n
# y toma time_steps datos.
# Comienza a avanzar hacia adelante.
for i in range(len(X) - time_steps):
v = X.iloc[i:(i + time_steps)].values
Xs.append(v)
ys.append(y.iloc[i + time_steps])
return np.array(Xs), np.array(ys)
time_steps = 50
# reshape to [samples, time_steps, n_features]
X_train, y_train = create_dataset(train, train, time_steps)
X_test, y_test = create_dataset(test, test, time_steps)
print("X_train.shape = ", X_train.shape)
print("y_train.shape = ", y_train.shape)
print("X_test.shape = ", X_test.shape)
print("y_test.shape = ", y_test.shape)
X_train.shape = (1410, 50, 1) y_train.shape = (1410, 1) X_test.shape = (316, 50, 1) y_test.shape = (316, 1)
fig = px.line(df_norm, x=df.index, y='ValorObservadoNormalizado')
fig.update_xaxes(title_text="Fecha")
fig.update_yaxes(title_text="ValorObservadoNormalizado")
fig.show()
# shapes
inputs_shape = (X_train.shape[1], X_train.shape[2])
LSTM_output = 60
# layers
inputs = L.Input(inputs_shape)
lstm = L.LSTM(units=LSTM_output, name='LSTM')(inputs)
outputs_LSTM = L.Dense(1)(lstm)
LSTM_model = Model(inputs=inputs, outputs=outputs_LSTM, name='series_LSTM_model')
# Compiling the RNN
LSTM_model.compile(optimizer=Adam(0.001),loss="mean_squared_error")
# Fitting to the training set
start = time.time()
LSTM = LSTM_model.fit(
X_train,
y_train,
epochs=50,
batch_size=16,
validation_split=0.1,
verbose=1,
shuffle=False
)
print("compilation time : ", time.time() - start)
Epoch 1/50 80/80 [==============================] - 5s 29ms/step - loss: 0.0319 - val_loss: 0.0048 Epoch 2/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0096 - val_loss: 0.0045 Epoch 3/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0092 - val_loss: 0.0043 Epoch 4/50 80/80 [==============================] - 2s 31ms/step - loss: 0.0088 - val_loss: 0.0041 Epoch 5/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0084 - val_loss: 0.0040 Epoch 6/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0081 - val_loss: 0.0038 Epoch 7/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0079 - val_loss: 0.0038 Epoch 8/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0076 - val_loss: 0.0037 Epoch 9/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0074 - val_loss: 0.0036 Epoch 10/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0072 - val_loss: 0.0035 Epoch 11/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0071 - val_loss: 0.0035 Epoch 12/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0069 - val_loss: 0.0035 Epoch 13/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0068 - val_loss: 0.0034 Epoch 14/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0066 - val_loss: 0.0034 Epoch 15/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0065 - val_loss: 0.0033 Epoch 16/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0064 - val_loss: 0.0033 Epoch 17/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0063 - val_loss: 0.0033 Epoch 18/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0062 - val_loss: 0.0033 Epoch 19/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0062 - val_loss: 0.0033 Epoch 20/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0061 - val_loss: 0.0032 Epoch 21/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0061 - val_loss: 0.0032 Epoch 22/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0060 - val_loss: 0.0032 Epoch 23/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0060 - val_loss: 0.0032 Epoch 24/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0059 - val_loss: 0.0032 Epoch 25/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0059 - val_loss: 0.0032 Epoch 26/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0059 - val_loss: 0.0032 Epoch 27/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0059 - val_loss: 0.0032 Epoch 28/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0058 - val_loss: 0.0032 Epoch 29/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0058 - val_loss: 0.0031 Epoch 30/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0058 - val_loss: 0.0031 Epoch 31/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0058 - val_loss: 0.0031 Epoch 32/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0058 - val_loss: 0.0031 Epoch 33/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0058 - val_loss: 0.0031 Epoch 34/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0058 - val_loss: 0.0031 Epoch 35/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 36/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 37/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 38/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 39/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 40/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 41/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 42/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 43/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 44/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 45/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 46/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 47/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 48/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 49/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 50/50 80/80 [==============================] - 2s 24ms/step - loss: 0.0057 - val_loss: 0.0031 compilation time : 94.12628173828125
joblib.dump(LSTM_model,'/content/drive/Shareddrives/Mineria /LSTM')
WARNING:absl:Found untraced functions such as lstm_cell_layer_call_fn, lstm_cell_layer_call_and_return_conditional_losses while saving (showing 2 of 2). These functions will not be directly callable after loading.
['/content/drive/Shareddrives/Mineria /LSTM']
LSTM_model.summary()
Model: "series_LSTM_model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 50, 1)] 0
LSTM (LSTM) (None, 60) 14880
dense (Dense) (None, 1) 61
=================================================================
Total params: 14,941
Trainable params: 14,941
Non-trainable params: 0
_________________________________________________________________
LSTM_losses = pd.DataFrame(LSTM.history)
fig = px.line(LSTM_losses, x=LSTM_losses.index, y=["loss", "val_loss"])
fig.update_xaxes(title_text="Epoch")
fig.update_yaxes(title_text="Loss")
fig.show()
LSTM_Predict = LSTM_model.predict(X_test)
LSTM_Predict = scaler.inverse_transform(LSTM_Predict)
10/10 [==============================] - 1s 16ms/step
# plot y_train, y_test, and testPredict using plotly
seq_len = 50
fig = go.Figure()
fig.add_trace(
go.Scatter(
x=df.index[seq_len : len(y_train) + seq_len],
y=scaler.inverse_transform(y_train).ravel(),
mode="lines",
name="Entrenamiento",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train) + seq_len :],
y=scaler.inverse_transform(y_test).ravel(),
mode="lines",
name="Prueba",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train) + seq_len :],
y=LSTM_Predict.ravel(),
mode="lines",
name="PredicciĆ³n",
)
)
fig.update_xaxes(title_text="Fecha")
fig.update_yaxes(title_text="ValorObservado")
fig.show()
#funcion de perdidad de quantiles
def QuantileLoss(perc, delta=1e-4):
perc = np.array(perc).reshape(-1)
perc.sort()
perc = perc.reshape(1, -1)
def _qloss(y, pred):
I = tf.cast(y <= pred, tf.float32)
d = K.abs(y - pred)
correction = I * (1 - perc) + (1 - I) * perc
# huber loss
huber_loss = K.sum(correction * tf.where(d <= delta, 0.5 * d ** 2 / delta, d - 0.5 * delta), -1)
# order loss
q_order_loss = K.sum(K.maximum(0.0, pred[:, :-1] - pred[:, 1:] + 1e-6), -1)
return huber_loss + q_order_loss
return _qloss
# quantiles
perc_points = [0.025, 0.975]
# shapes
inputs_shape = (X_train.shape[1], X_train.shape[2])
LSTM_output = 60
# layers
inputs = L.Input(inputs_shape)
qlstm = L.LSTM(units=LSTM_output, name='LSTM')(inputs)
qoutputs_LSTM = L.Dense(2)(qlstm)
qLSTM_model = Model(inputs=inputs, outputs=qoutputs_LSTM, name='series_LSTM_model')
# Compiling the RNN
qLSTM_model.compile(optimizer=Adam(0.001), loss=QuantileLoss(perc_points))
# Fitting to the training set
start = time.time()
qLSTM = qLSTM_model.fit(
X_train,
y_train,
epochs=50,
batch_size=16,
validation_split=0.1,
verbose=1,
shuffle=False
)
print("compilation time : ", time.time() - start)
Epoch 1/50 80/80 [==============================] - 5s 30ms/step - loss: 0.0646 - val_loss: 0.0092 Epoch 2/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0150 - val_loss: 0.0094 Epoch 3/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0143 - val_loss: 0.0089 Epoch 4/50 80/80 [==============================] - 2s 24ms/step - loss: 0.0142 - val_loss: 0.0090 Epoch 5/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0136 - val_loss: 0.0085 Epoch 6/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0134 - val_loss: 0.0088 Epoch 7/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0132 - val_loss: 0.0086 Epoch 8/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0132 - val_loss: 0.0090 Epoch 9/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0129 - val_loss: 0.0086 Epoch 10/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0127 - val_loss: 0.0086 Epoch 11/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0124 - val_loss: 0.0083 Epoch 12/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0124 - val_loss: 0.0083 Epoch 13/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0123 - val_loss: 0.0082 Epoch 14/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0122 - val_loss: 0.0081 Epoch 15/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0120 - val_loss: 0.0080 Epoch 16/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0120 - val_loss: 0.0084 Epoch 17/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0119 - val_loss: 0.0082 Epoch 18/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0118 - val_loss: 0.0083 Epoch 19/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0117 - val_loss: 0.0080 Epoch 20/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0116 - val_loss: 0.0079 Epoch 21/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0116 - val_loss: 0.0081 Epoch 22/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0114 - val_loss: 0.0079 Epoch 23/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0114 - val_loss: 0.0078 Epoch 24/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0112 - val_loss: 0.0074 Epoch 25/50 80/80 [==============================] - 3s 33ms/step - loss: 0.0110 - val_loss: 0.0073 Epoch 26/50 80/80 [==============================] - 3s 34ms/step - loss: 0.0110 - val_loss: 0.0074 Epoch 27/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0109 - val_loss: 0.0073 Epoch 28/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0109 - val_loss: 0.0074 Epoch 29/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0107 - val_loss: 0.0073 Epoch 30/50 80/80 [==============================] - 2s 24ms/step - loss: 0.0106 - val_loss: 0.0073 Epoch 31/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0106 - val_loss: 0.0072 Epoch 32/50 80/80 [==============================] - 2s 24ms/step - loss: 0.0105 - val_loss: 0.0072 Epoch 33/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0105 - val_loss: 0.0074 Epoch 34/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0104 - val_loss: 0.0075 Epoch 35/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0105 - val_loss: 0.0074 Epoch 36/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0104 - val_loss: 0.0074 Epoch 37/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0104 - val_loss: 0.0074 Epoch 38/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0104 - val_loss: 0.0074 Epoch 39/50 80/80 [==============================] - 2s 22ms/step - loss: 0.0104 - val_loss: 0.0074 Epoch 40/50 80/80 [==============================] - 2s 24ms/step - loss: 0.0104 - val_loss: 0.0074 Epoch 41/50 80/80 [==============================] - 3s 34ms/step - loss: 0.0104 - val_loss: 0.0074 Epoch 42/50 80/80 [==============================] - 3s 39ms/step - loss: 0.0104 - val_loss: 0.0073 Epoch 43/50 80/80 [==============================] - 3s 39ms/step - loss: 0.0103 - val_loss: 0.0073 Epoch 44/50 80/80 [==============================] - 2s 24ms/step - loss: 0.0103 - val_loss: 0.0073 Epoch 45/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0103 - val_loss: 0.0073 Epoch 46/50 80/80 [==============================] - 2s 31ms/step - loss: 0.0103 - val_loss: 0.0072 Epoch 47/50 80/80 [==============================] - 3s 34ms/step - loss: 0.0103 - val_loss: 0.0073 Epoch 48/50 80/80 [==============================] - 2s 24ms/step - loss: 0.0103 - val_loss: 0.0072 Epoch 49/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0104 - val_loss: 0.0073 Epoch 50/50 80/80 [==============================] - 2s 23ms/step - loss: 0.0102 - val_loss: 0.0071 compilation time : 144.6884093284607
qLSTM_model.summary()
Model: "series_LSTM_model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_2 (InputLayer) [(None, 50, 1)] 0
LSTM (LSTM) (None, 60) 14880
dense_1 (Dense) (None, 2) 122
=================================================================
Total params: 15,002
Trainable params: 15,002
Non-trainable params: 0
_________________________________________________________________
qLSTM_losses = pd.DataFrame(qLSTM.history)
fig = px.line(qLSTM_losses, x=qLSTM_losses.index, y=["loss", "val_loss"])
fig.update_xaxes(title_text="Epoch")
fig.update_yaxes(title_text="Loss")
fig.show()
qLSTM_Predict = qLSTM_model.predict(X_test)
qLSTM_Predict = scaler.inverse_transform(qLSTM_Predict)
10/10 [==============================] - 1s 10ms/step
# plot y_train, y_test, and testPredict using plotly
fig = go.Figure()
fig.add_trace(
go.Scatter(
x=df.index[seq_len : len(y_train) + seq_len],
y=scaler.inverse_transform(y_train).ravel(),
mode="lines",
name="Entrenamiento",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train) + seq_len :],
y=LSTM_Predict.ravel(),
mode="lines",
name="PredicciĆ³n",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train) + seq_len :],
y=qLSTM_Predict[:,0] ,
mode="lines",
name="0.025",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train) + seq_len :],
y=qLSTM_Predict[:,1],
mode="lines",
name="0.975",
)
)
fig.update_xaxes(title_text="Fecha")
fig.update_yaxes(title_text="ValorObservado")
fig.show()
LSTM_Score1 = metrics.mean_squared_error(scaler.inverse_transform(y_test), LSTM_Predict)
print('Test Score: %.2f RMSE' % (LSTM_Score1))
Test Score: 0.68 RMSE
def create_dataset(X, y, time_steps=1):
# crea dos listas vacias para depositar los datos
Xs, ys = [], []
# el primer lote de datos empieza en la primera observaciĆ³n
# y toma time_steps datos.
# Comienza a avanzar hacia adelante.
for i in range(len(X) - time_steps-5):
v = X.iloc[i:(i + time_steps)].values
Xs.append(v)
ys.append(y.iloc[i + time_steps+5])
return np.array(Xs), np.array(ys)
time_steps = 50
# reshape to [samples, time_steps, n_features]
X_train_5p, y_train_5p = create_dataset(train, train, time_steps)
X_test_5p, y_test_5p = create_dataset(test, test, time_steps)
print("X_train.shape = ", X_train_5p.shape)
print("y_train.shape = ", y_train_5p.shape)
print("X_test.shape = ", X_test_5p.shape)
print("y_test.shape = ", y_test_5p.shape)
X_train.shape = (1405, 50, 1) y_train.shape = (1405, 1) X_test.shape = (311, 50, 1) y_test.shape = (311, 1)
# shapes
inputs_shape = (X_train_5p.shape[1], X_train_5p.shape[2])
LSTM_output_5p = 60
# layers
inputs_5p = L.Input(inputs_shape)
lstm_5p = L.LSTM(units=LSTM_output_5p, name='LSTM')(inputs_5p)
outputs_LSTM_5p = L.Dense(1)(lstm_5p)
LSTM_model_5p = Model(inputs=inputs_5p, outputs=outputs_LSTM_5p, name='series_LSTM_model')
# Compiling the RNN
LSTM_model_5p.compile(optimizer=Adam(0.001),loss="mean_squared_error")
# Fitting to the training set
start = time.time()
LSTM_5p = LSTM_model_5p.fit(
X_train_5p,
y_train_5p,
epochs=50,
batch_size=16,
validation_split=0.1,
verbose=1,
shuffle=False
)
print("compilation time : ", time.time() - start)
Epoch 1/50 79/79 [==============================] - 3s 24ms/step - loss: 0.0261 - val_loss: 0.0060 Epoch 2/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0123 - val_loss: 0.0059 Epoch 3/50 79/79 [==============================] - 1s 17ms/step - loss: 0.0124 - val_loss: 0.0059 Epoch 4/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0124 - val_loss: 0.0058 Epoch 5/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0123 - val_loss: 0.0058 Epoch 6/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0123 - val_loss: 0.0058 Epoch 7/50 79/79 [==============================] - 1s 17ms/step - loss: 0.0122 - val_loss: 0.0057 Epoch 8/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0122 - val_loss: 0.0057 Epoch 9/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0121 - val_loss: 0.0057 Epoch 10/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0121 - val_loss: 0.0056 Epoch 11/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0121 - val_loss: 0.0056 Epoch 12/50 79/79 [==============================] - 1s 17ms/step - loss: 0.0120 - val_loss: 0.0056 Epoch 13/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0120 - val_loss: 0.0055 Epoch 14/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0120 - val_loss: 0.0055 Epoch 15/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0120 - val_loss: 0.0055 Epoch 16/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0120 - val_loss: 0.0055 Epoch 17/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0119 - val_loss: 0.0054 Epoch 18/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0119 - val_loss: 0.0054 Epoch 19/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0119 - val_loss: 0.0054 Epoch 20/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0119 - val_loss: 0.0054 Epoch 21/50 79/79 [==============================] - 1s 17ms/step - loss: 0.0119 - val_loss: 0.0054 Epoch 22/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0119 - val_loss: 0.0053 Epoch 23/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0119 - val_loss: 0.0053 Epoch 24/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0119 - val_loss: 0.0053 Epoch 25/50 79/79 [==============================] - 2s 24ms/step - loss: 0.0119 - val_loss: 0.0053 Epoch 26/50 79/79 [==============================] - 2s 27ms/step - loss: 0.0119 - val_loss: 0.0053 Epoch 27/50 79/79 [==============================] - 2s 23ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 28/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 29/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 30/50 79/79 [==============================] - 1s 17ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 31/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 32/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 33/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 34/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 35/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 36/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 37/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 38/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 39/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 40/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 41/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0118 - val_loss: 0.0053 Epoch 42/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0117 - val_loss: 0.0053 Epoch 43/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0117 - val_loss: 0.0053 Epoch 44/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0117 - val_loss: 0.0053 Epoch 45/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0117 - val_loss: 0.0053 Epoch 46/50 79/79 [==============================] - 1s 19ms/step - loss: 0.0117 - val_loss: 0.0054 Epoch 47/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0117 - val_loss: 0.0054 Epoch 48/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0117 - val_loss: 0.0054 Epoch 49/50 79/79 [==============================] - 2s 19ms/step - loss: 0.0117 - val_loss: 0.0054 Epoch 50/50 79/79 [==============================] - 1s 18ms/step - loss: 0.0117 - val_loss: 0.0055 compilation time : 74.54931282997131
joblib.dump(LSTM_model_5p,'/content/drive/Shareddrives/Mineria /LSTM_r')
WARNING:absl:Found untraced functions such as lstm_cell_2_layer_call_fn, lstm_cell_2_layer_call_and_return_conditional_losses while saving (showing 2 of 2). These functions will not be directly callable after loading.
['/content/drive/Shareddrives/Mineria /LSTM_r']
LSTM_model_5p.summary()
Model: "series_LSTM_model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_3 (InputLayer) [(None, 50, 1)] 0
LSTM (LSTM) (None, 60) 14880
dense_2 (Dense) (None, 1) 61
=================================================================
Total params: 14,941
Trainable params: 14,941
Non-trainable params: 0
_________________________________________________________________
LSTM_losses_5p = pd.DataFrame(LSTM_5p.history)
fig = px.line(LSTM_losses_5p, x=LSTM_losses_5p.index, y=["loss", "val_loss"])
fig.update_xaxes(title_text="Epoch")
fig.update_yaxes(title_text="Loss")
fig.show()
LSTM_Predict_5p = LSTM_model.predict(X_test_5p )
LSTM_Predict_5p = scaler.inverse_transform(LSTM_Predict_5p)
10/10 [==============================] - 0s 12ms/step
# plot y_train, y_test, and testPredict using plotly
seq_len = 50
fig = go.Figure()
fig.add_trace(
go.Scatter(
x=df.index[seq_len : len(y_train_5p) + seq_len],
y=scaler.inverse_transform(y_train_5p).ravel(),
mode="lines",
name="Entrenamiento",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train_5p) + seq_len :],
y=scaler.inverse_transform(y_test_5p).ravel(),
mode="lines",
name="Prueba",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train_5p) + seq_len :],
y=LSTM_Predict_5p.ravel(),
mode="lines",
name="PredicciĆ³n",
)
)
fig.update_xaxes(title_text="Fecha")
fig.update_yaxes(title_text="ValorObservado")
fig.show()
LSTM_Score2 = metrics.mean_squared_error(scaler.inverse_transform(y_test_5p), LSTM_Predict_5p)
print('Test Score: %.2f RMSE' % (LSTM_Score2))
Test Score: 1.49 RMSE
def create_dataset(X, y, time_steps=1):
# crea dos listas vacias para depositar los datos
Xs, ys = [], []
# el primer lote de datos empieza en la primera observaciĆ³n
# y toma time_steps datos.
# Comienza a avanzar hacia adelante.
for i in range(len(X) - time_steps):
v = X.iloc[i:(i + time_steps)].values
Xs.append(v)
ys.append(y.iloc[i + time_steps])
return np.array(Xs), np.array(ys)
time_steps1 = 100
# reshape to [samples, time_steps, n_features]
X_train_r, y_train_r = create_dataset(train, train, time_steps1)
X_test_r, y_test_r = create_dataset(test, test, time_steps1)
print("X_train.shape = ", X_train_r.shape)
print("y_train.shape = ", y_train_r.shape)
print("X_test.shape = ", X_test_r.shape)
print("y_test.shape = ", y_test_r.shape)
X_train.shape = (1360, 100, 1) y_train.shape = (1360, 1) X_test.shape = (266, 100, 1) y_test.shape = (266, 1)
# shapes
inputs_shape = (X_train_r.shape[1], X_train_r.shape[2])
LSTM_output_r = 60
# layers
inputs_r = L.Input(inputs_shape)
lstm_r = L.LSTM(units=LSTM_output_r, name='LSTM')(inputs_r)
outputs_LSTM_r = L.Dense(1)(lstm_r)
LSTM_model_r = Model(inputs=inputs_r, outputs=outputs_LSTM_r, name='series_LSTM_model')
# Compiling the RNN
LSTM_model_r.compile(optimizer=Adam(0.001),loss="mean_squared_error")
# Fitting to the training set
start = time.time()
LSTM_r = LSTM_model_r.fit(
X_train_r,
y_train_r,
epochs=50,
batch_size=16,
validation_split=0.1,
verbose=1,
shuffle=False
)
print("compilation time : ", time.time() - start)
Epoch 1/50 77/77 [==============================] - 6s 47ms/step - loss: 0.0333 - val_loss: 0.0044 Epoch 2/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0092 - val_loss: 0.0042 Epoch 3/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0088 - val_loss: 0.0040 Epoch 4/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0084 - val_loss: 0.0038 Epoch 5/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0081 - val_loss: 0.0037 Epoch 6/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0078 - val_loss: 0.0036 Epoch 7/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0076 - val_loss: 0.0035 Epoch 8/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0074 - val_loss: 0.0035 Epoch 9/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0072 - val_loss: 0.0034 Epoch 10/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0070 - val_loss: 0.0034 Epoch 11/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0069 - val_loss: 0.0033 Epoch 12/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0067 - val_loss: 0.0033 Epoch 13/50 77/77 [==============================] - 5s 71ms/step - loss: 0.0066 - val_loss: 0.0033 Epoch 14/50 77/77 [==============================] - 5s 62ms/step - loss: 0.0065 - val_loss: 0.0033 Epoch 15/50 77/77 [==============================] - 5s 61ms/step - loss: 0.0063 - val_loss: 0.0033 Epoch 16/50 77/77 [==============================] - 3s 45ms/step - loss: 0.0062 - val_loss: 0.0032 Epoch 17/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0062 - val_loss: 0.0032 Epoch 18/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0061 - val_loss: 0.0032 Epoch 19/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0060 - val_loss: 0.0032 Epoch 20/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0060 - val_loss: 0.0032 Epoch 21/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0059 - val_loss: 0.0032 Epoch 22/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0059 - val_loss: 0.0032 Epoch 23/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0059 - val_loss: 0.0032 Epoch 24/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0058 - val_loss: 0.0032 Epoch 25/50 77/77 [==============================] - 4s 48ms/step - loss: 0.0058 - val_loss: 0.0032 Epoch 26/50 77/77 [==============================] - 4s 58ms/step - loss: 0.0058 - val_loss: 0.0032 Epoch 27/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0058 - val_loss: 0.0031 Epoch 28/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0058 - val_loss: 0.0031 Epoch 29/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 30/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 31/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 32/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 33/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 34/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 35/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 36/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 37/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 38/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 39/50 77/77 [==============================] - 5s 63ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 40/50 77/77 [==============================] - 3s 43ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 41/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 42/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 43/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 44/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 45/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 46/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 47/50 77/77 [==============================] - 3s 43ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 48/50 77/77 [==============================] - 3s 42ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 49/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 Epoch 50/50 77/77 [==============================] - 3s 41ms/step - loss: 0.0057 - val_loss: 0.0031 compilation time : 204.54403972625732
joblib.dump(LSTM_model_5p,'/content/drive/Shareddrives/Mineria /LSTM_r')
WARNING:absl:Found untraced functions such as lstm_cell_2_layer_call_fn, lstm_cell_2_layer_call_and_return_conditional_losses while saving (showing 2 of 2). These functions will not be directly callable after loading.
['/content/drive/Shareddrives/Mineria /LSTM_r']
LSTM_model_r.summary()
Model: "series_LSTM_model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_4 (InputLayer) [(None, 100, 1)] 0
LSTM (LSTM) (None, 60) 14880
dense_3 (Dense) (None, 1) 61
=================================================================
Total params: 14,941
Trainable params: 14,941
Non-trainable params: 0
_________________________________________________________________
LSTM_losses_r = pd.DataFrame(LSTM_r.history)
fig = px.line(LSTM_losses_r, x=LSTM_losses_r.index, y=["loss", "val_loss"])
fig.update_xaxes(title_text="Epoch")
fig.update_yaxes(title_text="Loss")
fig.show()
LSTM_Predict_r = LSTM_model_r.predict(X_test_r)
LSTM_Predict_r = scaler.inverse_transform(LSTM_Predict_r)
9/9 [==============================] - 1s 14ms/step
# plot y_train, y_test, and testPredict using plotly
seq_len1 = 100
fig = go.Figure()
fig.add_trace(
go.Scatter(
x=df.index[seq_len1 : len(y_train_r) + seq_len1],
y=scaler.inverse_transform(y_train_r).ravel(),
mode="lines",
name="Entrenamiento",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train_r) + seq_len1 :],
y=scaler.inverse_transform(y_test_r).ravel(),
mode="lines",
name="Prueba",
)
)
fig.add_trace(
go.Scatter(
x=df.index[len(y_train_r) + seq_len1 :],
y=LSTM_Predict_r.ravel(),
mode="lines",
name="PredicciĆ³n",
)
)
fig.update_xaxes(title_text="Fecha")
fig.update_yaxes(title_text="ValorObservado")
fig.show()
LSTM_Score3 = metrics.mean_squared_error(scaler.inverse_transform(y_test_r), LSTM_Predict_r)
print('Test Score: %.2f RMSE' % (LSTM_Score3))
Test Score: 0.63 RMSE
print('Modelo LSTM 1 paso adelante 50 retardos %f' % LSTM_Score1)
print('Modelo LSTM 5 paso adelante 50 retardos %f' % LSTM_Score2)
print('Modelo LSTM 1 paso adelante 100 retardos %f' % LSTM_Score3)
#print( "ECM para modelo GRU: %f" % error3) #GRU
Modelo LSTM 1 paso adelante 50 retardos 0.680276 Modelo LSTM 5 paso adelante 50 retardos 1.494980 Modelo LSTM 1 paso adelante 100 retardos 0.633088